Abstract: An apparatus for determining emotional arousal of a subject by speech analysis, and an associated method. In the method, a speech sample is obtained, the speech sample is pre-processed into silent and active speech segments and the active speech segments are divided into strings of equal length blocks (the blocks having primary speech parameters including pitch and amplitude parameters), a plurality of selected secondary speech parameters indicative of characteristics of equal-pitch are derived, rising-pitch and falling-pitch trends in the strings of blocks, the secondary speech parameters are compared with predefined, subject independent values representing non-emotional speech to generate a processing result indicative of emotional arousal, and the generated processed result is outputted to an output device.

Abstract: Embodiments of methods and means for correcting auto-correlated wireless signal samples are provided. Such embodiments include isolating and subtracting an interference vector from auto-correlated signal samples so that a corrected signal sample data set is derived. The corrected signal samples are then used in detecting and identifying symbols within the original wireless signal. Reliable and expeditious wireless communications can be achieved in accordance with the present embodiments.

Abstract: An apparatus for linear predictive coding of an audio signal comprises a segmentation processor (201) which generates signal segments for the audio signal. An autocorrelation processor (401) for generates a first autocorrelation sequence for each signal segment and a modification processor (403) generates a second autocorrelation sequence for each signal segment by modifying the first autocorrelation sequence in response to at least one psychoacoustic characteristic. A prediction coefficient processor (405) determines linear predictive coding coefficients for each signal segment in response to the second autocorrelation sequence. The invention allows a low complexity linear encoding which takes into account psychoacoustic considerations thereby allowing an improved perceived coding quality for a given data rate.

Abstract: A method for lost speech samples recovery in speech transmission systems is disclosed. The method employs a waveform coder operating on digital speech samples. It exploits the composite model of speech, wherein each speech segment contains both periodic and colored noise components, and separately estimates these two components of the unreliable samples. First, adaptive FIR filters computed from received signal statistics are used to interpolate estimates of the periodic component for the unreliable samples. These FIR filters are inherently stable and typically short, since only strongly correlated elements of the signal corresponding to pitch offset samples are used to compute the estimate. These periodic estimates are also computed for sample times corresponding to reliable samples adjacent to the unreliable sample interval. The differences between these reliable samples and the corresponding periodic estimates are considered as samples of the noise component.

Abstract: A method for transcoding a bit stream encoded according to a linear predictive coding (LPC) standard to a bit stream encoded according to a mixed-excitation linear prediction (MELP) standard, including: decoding a bit stream into a first set of vocoder parameters compatible with the LPC standard; transforming the first set of vocoder parameters into a second set of vocoder parameters compatible with the MELP standard without converting the first set of vocoder parameters to an analog or digital waveform representation; and encoding the second set of vocoder parameters into a bit stream compatible with the MELP vocoder standard.

Abstract: A method and an apparatus accurately discriminates between speech and voice-band data (VBD) in a communication network by calculating self similarity ratio (SSR) values, which indicate periodicity characteristics of an input signal segment, and/or autocorrelation coefficients, which indicate spectral characteristics of an input signal segment, to generate a speech/VBD discrimination result. In one implementation, the speech-VBD discriminating apparatus calculates both short-term delay and long-term delay SSR values to analyze the repetition rate of an input signal frame, thereby indicating whether the input signal frame has the periodicity characteristics of a typical speech signal or a VBD signal. The speech-VBD discriminating apparatus further calculates a plurality of short-term autocorrelation coefficients to determine the spectral envelope of an input frame, thereby facilitating accurate speech/VBD discrimination.

Abstract: A speech analysis apparatus analyzing prosodic characteristics of speech information and outputting a prosodic discrimination result includes an input unit inputting speech information, an acoustic analysis unit calculating relative pitch variation and a discrimination unit performing speech discrimination processing, in which the acoustic analysis unit calculates a current template relative pitch difference, determining whether a difference absolute value between the current template relative pitch difference and a previous template relative pitch difference is equal to or less than a predetermined threshold or not, when the value is not less than the threshold, calculating an adjacent relative pitch difference, and when the adjacent relative pitch difference is equal to or less than a previously set margin value, executing correction processing of adding or subtracting an octave of the current template relative pitch difference to calculate the relative pitch variation by applying the relative pitch differe

Abstract: A speech classification technique for robust classification of varying modes of speech to enable maximum performance of multi-mode variable bit rate encoding techniques. A speech classifier accurately classifies a high percentage of speech segments for encoding at minimal bit rates, meeting lower bit rate requirements. Highly accurate speech classification produces a lower average encoded bit rate, and higher quality decoded speech. The speech classifier considers a maximum number of parameters for each frame of speech, producing numerous and accurate speech mode classifications for each frame. The speech classifier correctly classifies numerous modes of speech under varying environmental conditions.

Abstract: A device and a method for estimating an open-loop pitch in a general speech CODEC are disclosed. The open-loop pitch estimation device includes an autocorrelation function calculation unit which calculates a normalized autocorrelation function from a perceptual weighing filtered speech signal, a maximum autocorrelation function and lag estimation unit which estimates a maximum autocorrelation function and candidates for the maximum autocorrelation function, a pitch candidate decision unit which decides candidates for a pitch by using the ratio of the estimated maximum autocorrelation function to the candidates for the estimated maximum autocorrelation function, and lags of which values are smaller than a predetermined threshold value, and a pitch estimation unit which estimates a pitch between the candidates for a pitch and the lags corresponding to the estimated maximum autocorrelation function by using a pitch of a previous frame of the speech signal.

Abstract: There is provided a pitch lag predictor for use by a speech decoder to generate a predicted pitch lag parameter. The pitch lag predictor comprises a summation calculator configured to generate a first summation based on a plurality of previous pitch lag parameters, and a second summation based on a plurality of previous pitch lag parameters and a position of each of the plurality of previous pitch lag parameters with respect to the predicted pitch lag parameter; a coefficient calculator configured to generate a first coefficient using a first equation based on the first summation and the second summation, and a second coefficient using a second equation based on the first summation and the second summation, wherein the first equation is different than the second equation; and a predictor configured to generate the predicted pitch lag parameter based on the first coefficient and the second coefficient.

Abstract: An embodiment may include an apparatus comprising a dejitter buffer to receive packets containing audio data, a codec coupled with the dejitter buffer, the codec to receive coded audio frames from the dejitter buffer and decode them, and a concealed seconds meter coupled with the dejitter buffer, the concealed seconds meter to record concealment events by the decoder to provide an objective measure of media impairment. Another exemplary embodiment may be a method comprising receiving packets containing audio information at a dejitter buffer, decomposing the packets to coded audio frames, sending the coded audio frames to a decoder and decoding the frames, generating a concealment output stream if the decoder does not receive a valid frame from the dejitter buffer, and recording concealment events to provide an objective measure of media impairment.

Abstract: A lossless audio codec encodes/decodes a lossless variable bit rate (VBR) bitstream with random access point (RAP) capability to initiate lossless decoding at a specified segment within a frame and/or multiple prediction parameter set (MPPS) capability partitioned to mitigate transient effects. This is accomplished with an adaptive segmentation technique that fixes segment start points based on constraints imposed by the existence of a desired RAP and/or detected transient in the frame and selects a optimum segment duration in each frame to reduce encoded frame payload subject to an encoded segment payload constraint. In general, the boundary constraints specify that a desired RAP or detected transient must lie within a certain number of analysis blocks of a segment start point.

Abstract: A voice activated camera is described which allows users to take remote photographs by speaking one or more keywords. In a preferred embodiment, a speech processing unit is provided which is arranged to detect extended periodic signals from a microphone of the camera. A control unit is also provided to control the taking of a photograph when such an extended periodic component is detected by the speech processing unit.

Abstract: A digital signal processing method and learning method and devices therefor, and a program storage medium which are capable of further improving the waveform reproducibility of a digital signal. Self correlation coefficients are calculated by cutting parts out of the digital signal by multiple windows having different sizes, and the parts are classified based on the calculation results of the self correlation coefficients. Then, the digital signal is converted by the prediction method corresponding to the classified class, so that the conversion further suitable for the features of the digital signal can be conducted.

Abstract: In one embodiment, a block of audio data is partitioned into N sub-blocks, and a plurality of code parameters s(0), s(1), . . . s(N?1) are determined. The N sub-blocks are entropy encoded using a plurality of entropy codes defined by the plurality of code parameters, and the code parameters are transmitted. The code parameters may be transmitted by directly transmitting s(0) representing the code parameter of the first sub-block, and transmitting a difference s(i)?s(i?1) for i=1, . . . N?1, s(i) representing the code parameter of an ith sub-block following the first sub-block.

Abstract: A communication system includes a destination that receives voice samples and a voice parameter generated by a source. The destination uses the voice samples and voice parameter to reconstruct voice information in response to a packet loss. The destination may reconstruct voice information from multiple sources.

Abstract: Autocorrelation values are determined as a basis for an estimation of a pitch lag in a segment of an audio signal. A first considered delay range for the autocorrelation computations is divided into a first set of sections, and first autocorrelation values are determined for delays in a plurality of sections of this first set of sections. A second considered delay range for the autocorrelation computations is divided into a second set of sections such that sections of the first set and sections of the second set are overlapping. Second autocorrelation values are determined for delays in a plurality of sections of this second set of sections.

Abstract: A prosody identifying apparatus for identifying input speech on the basis of prosodic features of the input speech is provided. The prosody identifying apparatus includes a sound analyzing section for acquiring an amount of change in movement of a feature distribution obtained from an autocorrelation matrix of the frequency characteristic of the input speech and an identifying section for recognizing the input speech on the basis of an output of the sound analyzing section.

Abstract: A random code vector reading section and a random codebook of a conventional CELP type speech coder/decoder are respectively replaced with an oscillator for outputting different vector streams in accordance with values of input seeds, and a seed storage section for storing a plurality of seeds. This makes it unnecessary to store fixed vectors as they are in a fixed codebook (ROM), thereby considerably reducing the memory capacity.

Abstract: The present invention discloses a complete speech recognition system having a training button and a recognition button, and the whole system uses the application specific integrated circuit (ASIC) architecture for the design, and also uses the modular design to divide the speech processing into 4 modules: system control module, autocorrelation and linear predictive coefficient module, cepstrum module, and DTW recognition module. Each module forms an intellectual product (IP) component by itself. Each IP component can work with various products and application requirements for the design reuse to greatly shorten the time to market.

Abstract: An apparatus and a method for symbol decoding of baseband data in a wireless communications network is disclosed, and specifically CCK subsymbol prediction and symbol demodulation that occurs at 5.5 Mbps or 11 Mbps. The apparatus is configured to demodulate or predict the data differently, depending on the modulation rate. If the data was modulated at 11 Mbps, the ?3 rotator is rotated through each of its possible phase values and symbol correlation takes four clock cycles to complete. If the data was modulated at 5.5 Mbps, ?3 is not rotated with a set value of 0 within the correlator architecture, thereby saving power and reducing symbol correlation and subsymbol prediction to a single cycle while in such transmission mode.

Abstract: A method for detecting and attenuating inhalation noise in a communication system coupled to a pressurized air delivery system, the method including the steps of: generating an inhalation noise model (912, 1012) based on inhalation noise; receiving an input signal (802) that includes inhalation noise; comparing (810) the input signal to the noise model to obtain a similarity measure; determining (854) a gain factor based on the similarity measure; and modifying (852) the input signal based on the gain factor, wherein the inhalation noise in the input signal is attenuated based on the gain factor.

Abstract: A speech processing system is provided which is operable to receive sets of signal values representative of a speech signal generated by a speech source as distorted by a transmission channel between the speech source and the speech processing system. The system stores data defining a predetermined function derived from a signal model which models both the speech source and the channel and defining a probability density function which gives, for a given set of model parameters, the probability that the signal model has those model parameters given that the signal model is assumed to have generated the received set of signal values. The system applies a current set of received signal values to the stored probability density function and then draws samples from it using a Gibbs sampler. The system then analyses the samples to determine a set of parameter values representative of the speech signal before it was distorted by the channel.

Abstract: A system or method for modeling a signal, such as a speech signal, in which harmonic frequencies and amplitudes are identified and the harmonic magnitudes are interpolated to obtain spectral magnitudes at a set of fixed frequencies. An inverse transform is applied to the spectral magnitudes to obtain a pseudo auto-correlation sequence, from which linear prediction coefficients are calculated. From the linear prediction coefficients, model harmonic magnitudes are generated by sampling the spectral envelope defined by the linear prediction coefficients. A set of scale factors are then calculated as the ratio of the harmonic magnitudes to the model harmonic magnitudes and interpolated to obtain a second set of scale factors at the set of fixed frequencies. The spectral envelope magnitudes at the set of fixed frequencies are multiplied by the second set of scale factors to obtain new spectral magnitudes and the process is iterated to obtain final linear prediction coefficients.

Abstract: A communication system includes a destination that receives voice samples and a voice parameter generated by a source. The destination uses the voice samples and voice parameter to reconstruct voice information in response to a packet loss. The destination may reconstruct voice information from multiple sources.

Abstract: A reception AGC circuit includes a high-speed power calculating circuit for calculating the power of an input signal at a short period, a normal power calculating circuit for calculating the power at a normal period, a circuit for receiving a power calculation result from the high-speed power calculating circuit or normal power calculating circuit to calculate a feedback amplification value, and addition amplification value setting units for self-station communication and peripheral station monitoring which receive the feedback amplification value through a switch and add the feedback amplification value to an amplifier in use.

Abstract: After one or both of a pair of images are obtained, an auto-correlation function for one of those images is generated to determine a smear amount and possibly a smear direction. The smear amount and direction are used to identify potential locations of a peak portion of the correlation function between the pair of images. The pair of images is then correlated only at offset positions corresponding to the one or more of the potential peak locations. In some embodiments, the pair of images is correlated according to a sparse set of image correlation function value points around the potential peak locations. In other embodiments, the pair of images is correlated at a dense set of correlation function value points around the potential peak locations. The correlation function values of these correlation function value points are then analyzed to determine the offset position of the true correlation function peak.

Abstract: A method of estimating the pitch of a speech signal comprises the steps of sampling the speech signal to obtain a series of samples, dividing the series of samples into segments, each segment having a fixed number of consecutive samples, calculating for each segment a conformity function, and detecting peaks in the conformity function. The method provides also an intermediate signal derived from the speech signal, which is set to logical “1” where the intermediate signals exceeds a pre-selected threshold and to logical “0” where the intermediate signal does not exceed the pre-selected threshold, calculating the autocorrelation of the binary signal, and using the distance between peaks in the autocorrelation of the binary signal as an estimate of the pitch. Elaborate operations needed in prior art algorithms is thus avoided. A device conforming to the method is described.

Abstract: An application server generates and maintains a server-side data record, also referred to as a “brownie”, that includes application state information and user attribute information for multiple users within a single session controlled by a web-based browser. The brownie includes a session identifier that uniquely identifies the session, and a subsession identifier that uniquely identifies each corresponding user of the application session. As each new user is added to the session, for example by initiating a call to the new user, the application server stores the subsession identifier and corresponding application state information for the new user in the same brownie. In response to receiving a second web page request from the browser that includes the session identifier, the application server initiates a new web application instance, and recovers the brownie from the memory based on the session identifier included in the second page request.

Abstract: Partial waveform data representative of a waveform shape variation are extracted from supplied waveform data, and the extracted partial waveform data are stored along with time position information indicative of their respective time positions. In reproduction, the partial waveform data and time position information are read out, then the partial waveform data are arranged on the time axis in accordance with the time position information, and a waveform is produced on the basis of the waveform data arranged on the time axis. In another implementation, sets of sample identification information and time position information are obtained in accordance with a performance tone waveform to be reproduced, and sample data are obtained from a database in accordance with the sample identification information. The thus-obtained sample data are arranged on the time axis in accordance with the time position information, and the desired waveform is produced on the basis of the sample data arranged on the time axis.

Abstract: A prediction parameter analysis apparatus comprises a windowing part which generates a short time input signal by subjecting an input signal or a signal derived from the input signal to windowing, a component removal part which removes an unnecessary component from the short time input signal to generate a modified short time input signal, an autocorrelation coefficient computation part which computes autocorrelation coefficients based on the modified short time input signal, and a prediction parameter computation part which computes prediction parameters based on the autocorrelation coefficients.

Abstract: The invention concerns a method for conditioning a digital speech signal(s) processed by successive frames, which consists carrying out a harmonic analysis to estimate the pitch on each frame where it has a speech activity, and in oversampling at an oversampling frequency (fe) which is a multiple of the estimated pitch.

Abstract: A method and apparatus for providing fast data recovery with adaptive pulse code modulation (ADPCM) coding wherein, an ADPCM encoder periodically records the compressional parameters in memory together with regular compressed codes, and an ADPCM decoder retrieves the previously saved compressional parameters when reading the regular data from memory. In case any error occurs in the data compression and decompression processes which would cause a data divergence in the output of the ADPCM decoder, the previously saved compressional parameters can be used to correct the output data, thus enabling fast data recovery in the data outputting process without affecting downstream data in the data stream.

Abstract: An apparatus for analyzing an audio signal with regard to rhythm information of the audio signal comprises a filterbank for dividing the audio signal into at least two sub-band signals. Every sub-band signal is examined with regard to a periodicity of the sub-band signal to obtain rhythm raw-information of every sub-band signal. The rhythm raw-information is subjected to a quality evaluation to obtain a significance measure for every sub-band signal. The rhythm information of the audio signal will finally be established by considering the significance measure of the sub-band signal and the rhythm raw-information. This enables a more robust analysis of the audio signal, since sub-band signals, where significant rhythm information are present, are preferred compared to sub-band signals where less significant rhythm information are present, when establishing the rhythm information.

Abstract: A method and apparatus for classifying signals into a multiplicity of signal classes which employs discriminant functions of low-complexity discriminant variables that are computed directly from the passband signal. The method can be applied to the problem of classifying voiceband data (VBD), facsimile (FAX), native binary data, and speech on a 64 Kbps digital channel. In a hybrid two stage classification system, the first stage employs linear discriminant functions to make classification decisions into a smaller number of possible preliminary signal classes. The decisions of the first stage are then refined by a second stage that uses nonlinear discriminant functions such as quadratic or pseudo-quadratic functions. The second stage of a hybrid classifier then assigns the signal into a larger number of possible classes than does the first stage of the classifier alone.

Abstract: A voiced/unvoiced speech classifier (30) includes a speech segmentor (34) which segments an input digitized speech waveform into frames of speech and a band-pass filter (36) which filters the frames of speech. A relative energy generator (38) generates a relative energy value for each filtered frame of speech and a decision parameter generator (52) including an autocorrelation calculator (54) and a pitch calculator (56) generates a decision parameter based on an autocorrelation function and a pitch frequency index for the filtered frames of speech. A normalized energy calculator (46) adjusts the threshold and then normalizes the relative energy. A comparator (60) provides a signal indicative of whether a frame of speech is voiced speech or unvoiced speech depending on a comparison of the decision parameter and the normalized relative energy value for each filtered frame of speech.

Abstract: A system or method for modeling a signal, such as a speech signal, in which harmonic frequencies and amplitudes are identified and the harmonic magnitudes are interpolated to obtain spectral magnitudes at a set of fixed frequencies. An inverse transform is applied to the spectral magnitudes to obtain a pseudo auto-correlation sequence, from which linear prediction coefficients are calculated. From the linear prediction coefficients, model harmonic magnitudes are generated by sampling the spectral envelope defined by the linear prediction coefficients. A set of scale factors are then calculated as the ratio of the harmonic magnitudes to the model harmonic magnitudes and interpolated to obtain a second set of scale factors at the set of fixed frequencies. The spectral envelope magnitudes at the set of fixed frequencies are multiplied by the second set of scale factors to obtain new spectral magnitudes and the process is iterated to obtain final linear prediction coefficients.

Abstract: A voice conversion system employs a codebook mapping approach to transforming a source voice to sound like a target voice. Each speech frame is represented by a weighted average of codebook entries. The weights represent a perceptual distance of the speech frame and may be refined by a gradient descent analysis. The vocal tract characteristics, represented by a line spectral frequency vector, the excitation characteristics, represented by a linear predictive coding residual, the duration, and the amplitude of the speech frame are transformed in the same weighted-average framework.

Abstract: The present invention comprises a system and method for correlation domain formant enhancement. The method includes attenuating an output speech signal in areas of low spectral amplitude. The attenuated signals are subsequently analyzed, and then converted to a format suitable for transmission.

Abstract: The present invention comprises a system and method for frequency domain formant enhancement. The method includes attenuating an output speech signal in areas of low spectral amplitude. The attenuated signals are subsequently analyzed, and then converted to a format suitable for transmission.

Abstract: A method and system for injecting external content to a user's client computer engaged in an interactive computer network session. A request for selected content from a user is intercepted and a decision is made whether to deliver external content to the user's client computer in addition to the requested content. The method and system allows for local service providers such as ISPs to add their own content to sessions involving remote content suppliers.

Abstract: A method for the algebraic codebook search of a speech signal encoder, preferably using the Code Excited Linear Prediction process, in which, in order to calculate coefficients of the triangular matrix of the auto-correlation matrix of the Toeplitz type, a time interval comprising n speech signal samplings is divided into an integral number of tracks t with p possible pulse positions each, and in which the coefficients are stored in a memory grouped in combinations of adjacent tracks, combinations of non-adjacent tracks, combinations of identical tracks, and coefficients of the main diagonals of the auto-correlation matrix.

Abstract: A communications system (100) includes a multi-rate source coder (MRSC) (102), a variable size/rate buffer (VSRB) (112), a speech buffer (104), and a buffer control block (106). The variable size/rate buffer (112) includes a source coder bit buffer (SCBB) (114) and an adaptive transmit frame buffer (116). The source coder bit buffer (114) receives speech frames coded at different rates from the multi-rate source coder (102), and deposits an integer or non-integer number of frames in the adaptive transmit frame buffer (ATFB) (116). A receiver includes a seamless rate transition module (SRTM) (308) and an variable buffer (310). The seamless rate transition module (308) correlates speech data previously coded at different rates, and it then truncates or alternatively appends, concatenates, and warps the speech data to remove any annoying artifacts at the rate change boundary.

Abstract: A prediction parameter analysis apparatus comprises a windowing part which generates a short time input signal by subjecting an input signal or a signal derived from the input signal to windowing, a component removal part which removes an unnecessary component from the short time input signal to generate a modified short time input signal, an autocorrelation coefficient computation part which computes autocorrelation coefficients based on the modified short time input signal, and a prediction parameter computation part which computes prediction parameters based on the autocorrelation coefficients.

Abstract: An audio processing device includes an analyzer and a filter. The analyzer extracts an envelope of a noise signal and derives therefrom noise envelope parameters. The filter has coefficients which vary in response to noise envelope parameters and filters a useful signal to form a filtered signal. The coefficients are varied so that the filter enhances frequency bands of the useful signal that correspond to frequency bands of the noise signal having a higher energy than a predetermined value.

Abstract: A system determines a power spectral density associated with an audio signal that includes a speech signal and/or a noise signal. The system updates an autocorrelation function of the audio signal from samples in the audio signal, estimates an autocorrelation function of the speech signal from the updated autocorrelation function of the audio signal, and calculates a power spectral density of the speech signal using the estimated autocorrelation function. The system then determines the power spectral density of the audio signal from the calculated power spectral density of the speech signal.

Abstract: A pitch determiner (931) of a system controller (106) that generates a smoothed pitch value for a current frame of a low bit rate voice message includes a pitch function generator (955) that generates a pitch detection function (PDF) for each frame of digital samples of a voice signal, a pitch candidate selector (960) that selects a future frame pitch candidate from a pitch detection function (PDF), and a pitch adjuster (978) that generates the smoothed pitch value. The pitch adjuster includes a subharmonic pitch corrector (965) that determines a future frame pitch value by performing pitch subharmonic correction of a future frame pitch candidate using a roughness factor of the frequency transformed window.

Abstract: Bits are allocated to short-term repetition information for unvoiced input signals. Stated differently, more bits are allocated for pitch information during unvoiced input speech than in the prior art. The improved method and apparatus in an encoder (300) and decoder (700) result in improved consistency of amplitude pulses compared to prior art methods which indicates improved stability due to increased search resolution. Also, the improved method and apparatus result in higher energy compared to prior art methods which indicates that the synthesized waveform matches the target waveform more closely, resulting in a higher fixed codebook (FCB) gain.

Abstract: In a computerized method a continuous signal is segmented in order to determine statistically stationary units of the signal. The continuous signal is sampled at periodic intervals to produce a timed sequence of digital samples. Fixed numbers of adjacent digital samples are grouped into a plurality of disjoint sets or frames. A statistical distance between adjacent frames is determined. The adjacent sets are merged into a larger set of samples or cluster if the statistical distance is less than a predetermined threshold. In an iterative process, the statistical distance between the adjacent sets are determined, and as long as the distance is less than the predetermined threshold, the sets are iteratively merged to segment the signal into statistically stationary units.

Abstract: A pitch period extracting apparatus includes a microcomputer which determines a sampling frequency for an A/D converter, and a range of delay times for calculating autocorrelative values on the basis of the sampling frequency. For example, the delay times are set within a range of 20 samples?k?100 samples in a case of 8 kHz, and a range of 15 samples?k?75 samples in a case of 6 kHz. The microcomputer calculates the autocorrelative values of speech signal data stored in a buffer memory, and outputs a delay time at which a maximum autocorrelative value is obtainable as a pitch period of an inputted speech signal.